/systems-and-research

Engineering Systems & Research

Production deployments, hackathon-winning architectures, and empirical machine learning pipelines — spanning non-biometric edge inference, post-quantum cryptography discovery, biomimetic memory agents, and quantitative market microstructure.

KHOJ / Kho-Ya-Paya (Kumbh Mela 2027)
Inspect Full Screen
1st Place Winner · Claude Impact Lab Mumbai 2026

KHOJ / Kho-Ya-Paya (Kumbh Mela 2027)

Privacy-First Missing-Person Identification System

Winner at Claude Impact Lab Mumbai 2026. A non-biometric, edge-accelerated missing-person identification platform designed for 50-kiosk choke-point deployment at Kumbh Mela 2027 with zero facial recognition.

Edge AIMobileNetV3ResNet-50FastAPIVector SearchDPDP Act 2023Docker
Source Code
AdaptiveMarket — Self-Evolving Strategy Agent
Inspect Full Screen
Winning Submission · HackWithHyderabad 3.0

AdaptiveMarket — Self-Evolving Strategy Agent

Biomimetic Memory & Competitive Intelligence Engine

Winning submission at HackWithHyderabad 3.0. Autonomous competitive intelligence agent powered by Vectorize Hindsight 5-layer biomimetic memory and Groq LPU (openai/gpt-oss-120b) that mutates pricing, sales battlecards, and product roadmaps.

Next.js 15Vectorize HindsightGroq LPU (120B MoE)FastAPITEMPR RetrievalTypeScriptRender
Source CodeLive Deployment
Q-DARPAN (SIH 2026 / NTRO ECDAT)
Inspect Full Screen
SIH 2026 Finalist · NTRO ECDAT (Team PRAHARI)

Q-DARPAN (SIH 2026 / NTRO ECDAT)

Enterprise Cryptographic Discovery & Post-Quantum Risk

Selected for Smart India Hackathon (SIH 2026 / SIH26164 ECDAT) as Team PRAHARI. Generates CycloneDX 1.7 Cryptographic Bill of Materials (CBOM) and maps enterprise migration pathways to NIST Post-Quantum standards.

Post-Quantum CryptographyCycloneDX 1.7 CBOMAST ParsingPythonFastAPINIST PQC (FIPS 203/204)Docker
Source Code
ComplianceMind
Inspect Full Screen
Runner-Up Best Demo · YC Fall 2026 × Moss Sprint Bangalore

ComplianceMind

4-Agent Financial Compliance System (Y Combinator Fall 2026 × Moss Sprint)

Runner-Up for Best Demo & Presentation at the Y Combinator Fall 2026 × Moss Zero Latency Builder Sprint (Bangalore). A 4-agent system for financial compliance where agents communicate directly to detect short selling, fraud, and regulatory violations.

4-Agent Inter-CommunicationMoss Zero-Latency RetrievalNext.js 14TypeScriptFinancial Fraud DetectionFastAPIVercel
Source CodeLive Deployment
CKD Digital Twin
Inspect Full Screen
Live in Production · Team Hare

CKD Digital Twin

Chronic Kidney Disease Policy Simulator & Patient Risk Predictor

Healthcare digital twin and clinical simulation platform built by Team Hare. Simulates patient-specific disease progression trajectories and models regional health policy interventions.

Next.jsTypeScriptMachine LearningBiomarker ModelingTailwind CSSRechartsVercel
Source CodeLive Deployment
Campus Brain Android (Offline RAG)
Inspect Full Screen
88.9% QA Benchmark · 100% Offline

Campus Brain Android (Offline RAG)

Zero-Network On-Device SLM Knowledge Router

Four-route hierarchical retrieval engine (FACT/LOCAL/GLOBAL/TABULAR) achieving 88.9% benchmark accuracy, running entirely on-device on Android with native Kotlin and local SLMs.

KotlinAndroid SDKOn-Device SLMDuckDBVector SearchPythonFastAPI
Source Code
VishwaGuru
Inspect Full Screen
15+ Stars · GSSoC & ECWoC Project Admin

VishwaGuru

Open-Source AI Civic Governance & Grievance Platform

Open-source civic platform empowering Indian youth in democratic processes. Multilingual AI auto-routes citizen issues to local representatives; 15+ GitHub stars, Project Admin across GSSoC & ECWoC.

PythonReactFastAPINLPGeolocation APIsDockerGitHub Actions
Source Code
Kaggle Competitive ML & ARC Prize 2026
Inspect Full Screen
Top 14 & Top 20 Competitive ML

Kaggle Competitive ML & ARC Prize 2026

Top 14 & Top 20 Global Competitive Machine Learning Stacks

Top 14 finish in Kaggle Playground Series S6E10 (0.96225 Public LB), Top 20 competitive engine in Kaggle 'Name That LLM' with 24-LoRA ensembles, and symbolic DSL solver for ARC-AGI-2.

PythonCatBoostLightGBMXGBoostPyTorchLoRA EnsemblesARC-AGI-2 DSL
Source Code
Quantensor (Regime & Microstructure Study)
Inspect Full Screen
Research Manuscript · arXiv & IEEE CIFEr

Quantensor (Regime & Microstructure Study)

Fixed-Band Neutrality Collapse & Market Regime Modeling

Quantitative finance research & ML platform proving the h^(-1/2) neutrality collapse law and detecting regime shifts across 48 NSE equities and EURUSD. Manuscript prepared for arXiv & IEEE CIFEr.

PythonFastAPIDuckDBXGBoostAngel One SmartAPILaTeX / ResearchMarket Microstructure
Source Code
◊

Production & Research Systems

9

★

Claude Impact Lab Mumbai 2026

1st Place

▪

Kaggle S6E10 5-Fold CV ROC-AUC

0.96225